Would be interesting if smarter models are incoherent because they know more than us. But probably not there yet. https://x.com/anthropicai/status/2018481226999640355?s=46